Automating the categorisation of child sexual abuse images and videos, to protect victims faster, safeguard investigator mental wellbeing and reduce case time.
One of the most daunting parts of an investigation is the starting point at which an investigator needs to categorise images and videos, as a Canadian Digital Forensics ICE investigator shared with us.
After seizing, imaging and processing a range of phones, tablets and laptops for a CSAM investigation using various other software, all media was then ingested into S21 LASERi-X and it was ready for categorisation within seconds. S21 LASERi-X merged all media into a single analysis, de-duplicating content based on MD5 hashes but still keeping an audit trail of the media source device and individual metadata. The media was also pre-categorised based on MD5/PhotoDNA from the S21 Intelligence Database and Project VIC.
From the front screen of S21 LASERi-X, the investigator selected their preferred processing options along with the checkbox for ‘S21 Auto-Categoriser’ — which is now a key weapon in their arsenal. S21 LASERi-X imported all 900,000 unique files and duplicates in a little over a minute, and then started the S21 Auto-Categoriser in the background allowing the user to continue reviewing the media, running preliminary checks for local abuse victims, and categorising as usual.
The S21 Auto-Categoriser predicted for each piece of media whether it was CSAM, CSAM indicative, adult pornography or other; presenting its results in real-time to the investigator, with all images and videos completed in 80 minutes.
Using the S21 Auto-Categoriser prediction results, the investigator could quickly filter based on the predicted categories, confirming or adjusting the prediction result. They were essentially grouping files together based on the severity, with high accuracy and speed. This process allowed a more linear approach to categorisation, as opposed to the usual scatter-gun approach. The investigator reported the categorisation time was reduced by as much as 60% compared to their previous case work which had been completed without the S21 Auto-Categoriser.
In addition to the time savings, and the potential to find CSAM media faster, the investigator highlighted that they felt the “load had been lightened” and that they felt better, working alongside the S21 Auto-Categoriser than without it.